Airebo giving too large lattice parameter for graphite

I am using the AIREBO potential to simulate graphite (5x5x5 unit cells) with a LAMMPS version from 21Jun12. When I try to find the equilibrium lattice parameter through energy minimization, I get an in-plane lattice parameter of 1.61 Å even though it should be 1.396 for this potential. Varying the lattice parameter and calculating the energy to get an energy vs parameter curve gives the same result (total energy lowest at 1.61). Switching off the LJ and torsion terms gives an in-plane parameter of 1.398, so I suspect the problem is in the interplanar interactions, but I am using the recommended LJ cutoff of 3*sigma. Varying the cutoff does not help, neither does changing to 10x10x10 unit cells. Please help me spot the error!

Input file:
units metal
boundary p p p

atom_style atomic

#Positions input file
read_data data.graphite

#Force field
pair_style airebo 3.0 1 1
pair_coeff * * CH.airebo C

fix 1 all box/relax tri 0.0
min_style cg
min_modify line quadratic
minimize 1e-20 1e-10 10000 100000

write_restart test1.restart

The configuration file data.graphite is too long to post here, but it’s ordinary graphite with lattice parameters given by experiments (a=1.42, c=3.354). The cell looks like this:
0.0 12.295 xlo xhi
0.0 10.6477823395 ylo yhi
0.0 16.77 zlo zhi
-6.1475 0.0 0.0 xy xz yz

Best regards,

Edit Helgee, Ph.D. student
Department of Applied Physics
Chalmers Univ. of Technology
S-412 96 Goteborg, Sweden
Tel: +46 31 7723195

Aidan can comment, but another question on potentials and lattice constants
was answered last week by noting there is no guarantee that a particular
potential will reproduce a lattice constant, particularly for a material
with many phases. If there is a paper showing AIREBO gives the correct
graphite lattice constant, you can try to reproduce that calculation.